Integrate Generated Personalized Highlight Videos for Google Photos

Problem Statement

• Google Photos wants to generate personalized highlights videos. This requires:

• Pulling user data (photos, videos, metadata).

• Running ML jobs (video summarization).

• Delivering results to multiple frontends (mobile, web, Chromecast).

Questions to solve:

• How would you integrate batch ML jobs with real-time APIs?

• What’s your storage choice for intermediate results?

• How do you orchestrate workflows (Airflow vs event-driven)?

• How do you handle large-scale failures in the ML pipeline?